Scan2Score

Overview of the project

Inspiration

Evaluating exam papers manually is time-consuming and inconsistent. We wanted a tool that could digitize answer sheets instantly and provide meaningful feedback without requiring manual correction. Scan2Score was built to reduce effort while increasing fairness and clarity.

What it does

Scan2Score converts scanned question papers and answer scripts into digital text using OCR. It analyzes each student response using an AI model and generates clear, constructive feedback on how well the student addressed the question.

How we built it

We used PyMu PDF for PDF-to-image conversion and an LLM to interpret questions and evaluate student answers. The system follows a structured pipeline: OCR → text processing → AI evaluation → feedback generation.

Challenges we ran into

OCR accuracy varied across different scans, fonts, and formats. Ensuring that each answer aligned with the correct question required careful text segmentation. Another challenge was designing prompts that produced consistent, unbiased feedback from the LLM.

Accomplishments that we're proud of

We built a fully automated flow that can extract, analyze, and evaluate exam responses from a simple PDF upload. The system produces clean, insightful feedback without needing exact keywords or templates — just genuine understanding.

What we learned

We gained deeper insight into real-world OCR limitations, prompt engineering, and building reliable AI workflows. We also learned how to balance automation with clarity so the feedback remains accurate, useful, and student-friendly.

What's next for Scan2Score

We plan to add support for handwritten scripts, richer analytics for teachers, and a simple web dashboard for students. Long-term goals include multi-language support and integration with school learning platforms.